nvmf: add save/restore a controller in runtime

When doing live migration, there are some spdk_nvmf_ctrlr internal
data structures need to be saved/restored.

Change-Id: Ie39482e8c49765c36fc3700fbac4ce47ef306f29
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10058
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>

lib/nvmf/ctrlr.c

@@ -1922,6 +1922,51 @@ nvmf_ctrlr_save_aers(struct spdk_nvmf_ctrlr *ctrlr, uint16_t *aer_cids,
 	return ctrlr->nr_aer_reqs;
 }
 
+int
+nvmf_ctrlr_save_migr_data(struct spdk_nvmf_ctrlr *ctrlr, struct nvmf_ctrlr_migr_data *data)
+{
+	uint32_t num_async_events = 0;
+	struct spdk_nvmf_async_event_completion *event, *event_tmp;
+
+	memcpy(&data->feat, &ctrlr->feat, sizeof(struct spdk_nvmf_ctrlr_feat));
+	data->cntlid = ctrlr->cntlid;
+	data->acre_enabled = ctrlr->acre_enabled;
+	data->notice_aen_mask = ctrlr->notice_aen_mask;
+
+	STAILQ_FOREACH_SAFE(event, &ctrlr->async_events, link, event_tmp) {
+		data->async_events[num_async_events++].raw = event->event.raw;
+		if (num_async_events == NVMF_MIGR_MAX_PENDING_AERS) {
+			SPDK_ERRLOG("%p has too many pending AERs\n", ctrlr);
+			break;
+		}
+	}
+	data->num_async_events = num_async_events;
+
+	return 0;
+}
+
+int
+nvmf_ctrlr_restore_migr_data(struct spdk_nvmf_ctrlr *ctrlr, struct nvmf_ctrlr_migr_data *data)
+{
+	struct spdk_nvmf_async_event_completion *event;
+	uint32_t i;
+
+	memcpy(&ctrlr->feat, &data->feat, sizeof(struct spdk_nvmf_ctrlr_feat));
+	ctrlr->acre_enabled = data->acre_enabled;
+	ctrlr->notice_aen_mask = data->notice_aen_mask;
+
+	for (i = 0; i < data->num_async_events; i++) {
+		event = calloc(1, sizeof(struct spdk_nvmf_async_event_completion));
+		if (!event) {
+			return -ENOMEM;
+		}
+		event->event.raw = data->async_events[i].raw;
+		STAILQ_INSERT_TAIL(&ctrlr->async_events, event, link);
+	}
+
+	return 0;
+}
+
 static int
 nvmf_ctrlr_set_features_async_event_configuration(struct spdk_nvmf_request *req)
 {

lib/nvmf/nvmf_internal.h

@@ -282,6 +282,29 @@ struct spdk_nvmf_ctrlr {
 	TAILQ_ENTRY(spdk_nvmf_ctrlr)	link;
 };
 
+/* Maximum pending AERs that can be migrated */
+#define NVMF_MIGR_MAX_PENDING_AERS 256
+
+/* spdk_nvmf_ctrlr private migration data structure used to save/restore a controller */
+struct nvmf_ctrlr_migr_data {
+	uint32_t				opts_size;
+
+	uint16_t				cntlid;
+	uint8_t					reserved1[2];
+
+	struct spdk_nvmf_ctrlr_feat		feat;
+	uint32_t				reserved2[2];
+
+	uint32_t				num_async_events;
+	uint32_t				acre_enabled;
+	uint64_t				notice_aen_mask;
+	union spdk_nvme_async_event_completion	async_events[NVMF_MIGR_MAX_PENDING_AERS];
+
+	/* New fields shouldn't go after reserved3 */
+	uint8_t					reserved3[3000];
+};
+SPDK_STATIC_ASSERT(sizeof(struct nvmf_ctrlr_migr_data) == 0x1000, "Incorrect size");
+
 #define NVMF_MAX_LISTENERS_PER_SUBSYSTEM	16
 
 struct spdk_nvmf_subsystem {
@@ -452,6 +475,9 @@ void nvmf_ctrlr_abort_aer(struct spdk_nvmf_ctrlr *ctrlr);
 int nvmf_ctrlr_save_aers(struct spdk_nvmf_ctrlr *ctrlr, uint16_t *aer_cids,
 			 uint16_t max_aers);
 
+int nvmf_ctrlr_save_migr_data(struct spdk_nvmf_ctrlr *ctrlr, struct nvmf_ctrlr_migr_data *data);
+int nvmf_ctrlr_restore_migr_data(struct spdk_nvmf_ctrlr *ctrlr, struct nvmf_ctrlr_migr_data *data);
+
 /*
  * Abort zero-copy requests that already got the buffer (received zcopy_start cb), but haven't
  * started zcopy_end.  These requests are kept on the outstanding queue, but are not waiting for a

lib/nvmf/vfio_user.c

@@ -150,7 +150,7 @@ struct nvme_migr_device_state {
 
 	uint16_t	reserved2[3];
 	uint16_t	nr_aers;
-	uint16_t	aer_cids[256];
+	uint16_t	aer_cids[NVMF_MIGR_MAX_PENDING_AERS];
 
 	/* Controller private data offset and length if exist, starting at
 	 * the beginning of this data structure.
@@ -390,6 +390,7 @@ vfio_user_migr_data_len(void)
 	len = NVMF_VFIO_USER_MAX_QPAIRS_PER_CTRLR * (sizeof(struct nvme_migr_sq_state) + sizeof(
 				struct nvme_migr_cq_state));
 	len += sizeof(struct nvme_migr_device_state);
+	len += sizeof(struct nvmf_ctrlr_migr_data);
 	len += NVME_REG_BAR0_SIZE;
 	len += NVME_REG_CFG_SIZE;
 	/* BAR4 */